MBE Advance Access published November 24, 2015 ...chang.eeb.utoronto.ca/files/2016/02/Dungan-et-al-2016...Article Spectral Tuning of Killer Whale (Orcinus orca) Rhodopsin: Evidence

Article

Spectral Tuning of Killer Whale (Orcinus orca) Rhodopsin:Evidence for Positive Selection and Functional Adaptation in aCetacean Visual PigmentSarah Z. Dungan,1 Alexander Kosyakov,2 and Belinda S.W. Chang*,1,2,3

1Department of Ecology and Evolutionary Biology, University of Toronto, Toronto, ON, Canada2Department of Cell and Systems Biology, University of Toronto, Toronto, ON, Canada3Centre for the Analysis of Genome Evolution and Function, Toronto, ON, Canada

*Corresponding author: E-mail: [email protected].

Associate editor: James McInerney

Abstract

Cetaceans have undergone a remarkable evolutionary transition that was accompanied by many sensory adaptations,including modification of the visual system for underwater environments. Recent sequencing of cetacean genomes hasmade it possible to begin exploring the molecular basis of these adaptations. In this study we use in vitro expressionmethods to experimentally characterize the first step of the visual transduction cascade, the light activation of rhodopsin,for the killer whale. To investigate the spectral effects of amino acid substitutions thought to correspond with absorbanceshifts relative to terrestrial mammals, we used the orca gene as a background for the first site-directed mutagenesisexperiments in a cetacean rhodopsin. The S292A mutation had the largest effect, and was responsible for the majority ofthe spectral difference between killer whale and bovine (terrestrial) rhodopsin. Using codon-based likelihood models, wealso found significant evidence for positive selection in cetacean rhodopsin sequences, including on spectral tuning siteswe experimentally mutated. We then investigated patterns of ecological divergence that may be correlated with rho-dopsin functional variation by using a series of clade models that partitioned the data set according to phylogeny, habitat,and foraging depth zone. Only the model partitioning according to depth was significant. This suggests that foragingdives might be a selective regime influencing cetacean rhodopsin divergence, and our experimental results indicate thatspectral tuning may be playing an adaptive role in this process. Our study demonstrates that combining computationaland experimental methods is crucial for gaining insight into the selection pressures underlying molecular evolution.

Key words: cetacean vision, evolution of protein structure and function, codon substitution model, dN/dS, clade model,adaptive evolution, opsins, visual ecology, absorption spectra, dim-light vision, site-directed mutagenesis.

IntroductionCetaceans are the only mammals that have evolved a fullyaquatic lifestyle and diversified into all major aquatic ecosys-tems, a remarkable transition that has been characterized byextreme morphological and physiological adaptations (Uhen2010; Gatesy et al. 2013). The molecular mechanisms under-lying these major evolutionary changes are only just beginningto be understood, largely due to advances in cetacean whole-genome sequencing projects (Lindblad-Toh et al. 2011; Sunet al. 2013; Zhou et al. 2013; Moura et al. 2014; Yim et al. 2014).Recent studies that exploit cetacean sequence data have cor-related large-scale patterns of positive selection and gene losswith key aquatic adaptations including limb development,diving physiology, echolocation, and dim-light vision(Gatesy et al. 2013; McGowen et al. 2014). Because light de-tection is often essential for behaviours such as predatoravoidance, mate selection, and foraging, its evolution hasbeen directly linked to survivorship and reproductive fitness(Nilsson 2013). Marine environments present a particularchallenge to vision because light intensity is reduced, andthe spectrum is narrowed and blue-shifted relative to terres-trial environments (Warrant and Locket 2004). Compared

with most terrestrial mammals (including their nearestliving relatives, the hippopotamids) cetaceans have a reducednumber of visual pigments (reviewed in Bowmaker 2008 andJacobs 2013), and the ones they do possess show evidence ofspectral shifts (Fasick et al. 1998; Fasick and Robinson 2000;Newman and Robinson 2005). Nevertheless, the structure–function mechanisms responsible for these differences havenot been experimentally investigated in an evolutionary con-text in any cetacean pigment.

The killer whale (Orcinus orca) is one of the best-studiedcetacean species, making it a model for toothed cetacean(Odontoceti) biology and behavior in general (Baird 2000).It is the largest member of Delphinidae, and though it has acosmopolitan distribution in the world’s oceans, unlike otherwidely distributed delphinids, it tends not to venture beyondthe continental shelves into open ocean habitats (Wang et al.2014). Most foraging takes place in epipelagic waters where itis an apex predator, with over 140 species documented asprey including fish, cephalopods, marine mammals, seabirds,and marine turtles (Wang et al. 2014). In other cetaceans,foraging has likely driven several instances of ecological con-vergence, such as the evolution of bathypelagic foraging

� The Author 2015. Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution. All rights reserved. For permissions, pleasee-mail: [email protected]

Mol. Biol. Evol. doi:10.1093/molbev/msv217 Advance Access publication October 20, 2015 1

MBE Advance Access published November 24, 2015 at U

niversity of Toronto L

ibrary on January 17, 2016http://m

be.oxfordjournals.org/D

ownloaded from

http://mbe.oxfordjournals.org/

(below 1,000 m) in the sperm whales and beaked whales(Tyack et al. 2006; Watwood et al. 2006), and independentfreshwater invasions by the river dolphin clades (Cassens et al.2000; Hamilton et al. 2001). Though the killer whale ge-nome has recently become publicly available (Moura et al.2014), its visual pigments have yet to be experimentallyinvestigated.

Visual pigments are molecular complexes consisting of anopsin apoprotein covalently bound through a Schiff base link-age to a vitamin A-derived chromophore, 11-cis retinal (Wald1968; Sakmar et al. 1989; Palczewski et al. 2000). Visual pig-ments are activated when incoming photons trigger isomer-ization of 11-cis retinal to all-trans retinal, which results in aconformational change in the associated heptahelical opsinstructure, giving rise to metarhodopsin II, the biologicallyactive form (Lamb and Pugh 2004). Visual pigment moleculesare contained within the outer segments of photoreceptorcells, and are the first component of the visual transductionpathway. Similar to other mammals, cetaceans are known topossess three major classes of visual pigments: A rhodopsin(RH1) that mediates dim-light vision, and two classes of coneopsins (LWS and SWS1) that mediate daylight and colorvision, though in all cetaceans that have been analyzed sofar, SWS1 has been found to be pseudogenized (Bowmaker2008; Meredith et al. 2013).

Adaptive shifts in the peak wavelength of maximal absor-bance (�max) of visual pigments have been proposed toevolve in response to light environment, particularly in aqua-tic and nocturnal organisms (e.g., Bowmaker 2008; Hunt et al.2009; Jacobs 2009). Prior studies of cetacean pigmentshave found that they tend to be blue-shifted relative toterrestrial mammals, with the greatest shifts occurring indeep-diving species (McFarland 1971; Fasick et al. 1998;Fasick and Robinson 2000). Substitutions at three specificamino acid sites (83, 292, and 299) are thought to be largelyresponsible for these shifts (Fasick and Robinson 1998), butthese have yet to be experimentally mutated in a cetaceanpigment.

In this study, we have used in vitro expression and site-directed mutagenesis experiments to investigate spectraltuning in the rhodopsin from the killer whale. To determinethe potential for adaptive molecular evolution in cetaceanrhodopsin, we combined our experimental assays of killerwhale rhodopsin with computational investigations of selec-tion patterns in cetacean rhodopsin genes and their associa-tion with ecological variables such as foraging depth. Wehypothesized that these ecological differences among ceta-ceans may be driving divergent evolution in rhodopsin. Weinvestigated this hypothesis using codon-based random sitesand clade models to test for positive selection and divergencein cetacean rhodopsin evolution. Combined with our in vitroexperiments measuring the functional consequences ofamino acid substitutions at spectral tuning sites, our insilico analyses support the hypothesis that cetacean rhodop-sins may have been the targets of positive selection in order toadapt to changes in aquatic light environments, particularly inresponse to foraging depth.

Results

Killer Whale Rhodopsin Expression and Site-DirectedMutagenesis

Like most cetaceans, the killer whale possesses an intact genecoding for rhodopsin (RH1), although some of its other visualpigments, such as SWS1, may have been pseudogenized(Moura et al. 2014). We used heterologous expression meth-ods to express and characterize the absorption maximum(�max) of the wild-type killer whale rhodopsin (Morrowand Chang 2010). The expressed, purified rhodopsin wasfound to have a �max that was 12 nm blue-shifted relativeto the bovine control (fig. 1A and table 1). The killer whalepigment was also stable in hydroxylamine, similar to bovinerhodopsin (Sakmar et al. 1989; Kawamura and Yokoyama1998) (supplementary fig. S1, Supplementary Materialonline). Acid bleaching of killer whale rhodopsin shifted its�max to 440 nm (fig. 1B), indicative of an intact covalentlinkage with the retinal chromophore (Kito et al. 1968). AllRH1 pigments were capable of being bleached with light toapproximately 380 nm, indicating the formation of the light-activated state, metarhodopsin II (fig. 1A).

We then used site-directed mutagenesis to investigate theeffects of mutating spectral tuning residues 83, 292, and 299on the �max of cetacean rhodopsin. Single mutations of eachof these sites to the bovine rhodopsin residue resulted in clearspectral shifts (fig. 1C). S292A resulted in the largest red-shift(10 nm), with N83D resulting in a small red-shift (2 nm), andS299A in a small blue-shift (2 nm). A mutant with both red-shifting substitutions (N83D+S292A) was found to have syn-ergistic effects that resulted in a large red-shift (14 nm) (fig. 1Cand table 1).

To gain insight into the mechanisms by which the substi-tutions at these sites may influence �max, we constructed ahomology model of the wild-type killer whale rhodopsinbased on the dark state bovine rhodopsin crystal structure(Palczewski et al. 2000). Overall, the structure of killer whalerhodopsin was very similar to bovine rhodopsin (fig. 2A). Sites292 and 299 on helix VII were clustered near the chromo-phore and counterion, within 10 A of the Schiff base (though292 was closer at ~5 A) (fig. 2B–E). The serine substitutions atthese sites in killer whale, unlike the alanine residues in bovine,result in polar side chains in close proximity to the protonatedSchiff base end of the chromophore. Site 83 is located fartheraway (410 A) from the Schiff base on helix II, and is highlyconserved as aspartic acid (D) in terrestrial vertebrate rho-dopsins and other G protein-coupled receptors (GPCRs)where it participates in a hydrogen bond network withother residues that may influence chromophore activation(Okada et al. 2002). Though the N83 substitution in killerwhale is not charged, the configuration of this network ap-pears comparable to the bovine rhodopsin dark-state (fig. 2F).

Molecular Evolutionary Analyses

To investigate patterns of selection in cetacean rhodop-sin, we used codon-based likelihood models (reviewed inAnisimova and Koisol 2009) to estimate dN/dS for a data

2

Dungan et al. . doi:10.1093/molbev/msv217 MBE at U




ownloaded from

http://mbe.oxfordjournals.org/lookup/suppl/doi:10.1093/molbev/msv217/-/DC1



set of cetacean rhodopsin coding sequences.An alignment (supplementary fig. S3, SupplementaryMaterial online) of full rhodopsin coding sequencesfrom 23 cetacean species (supplementary table S1,Supplementary Material online) was analyzed both withrandom sites models implemented in PAML (Yang 2007),as well as models that independently estimate rates of dS

and dN in HYPHY (Pond et al. 2005). We found significantevidence not only for variation in the nonsynonymous tosynonymous substitution rate ratio, ! (dN/dS; M3 vs. M0,P< 0.0001; table 2 and supplementary table S2,Supplementary Material online) but also for pervasive

Nor

mal

ized

Abs

orba

nce

0

1

2

3

Wavelength (nm)

300 400 500 600

Killer whaleBovine

Light

487 nm 499 nm

Light bleach difference spectraA

489 nm

Nor

mal

ized

Abs

orba

nce

0

1

2

400 500 600

497 nm0

1

2

400 500 600

485 nm0

1

2

Wavelength (nm)400 500 600

501 nm0

1

2

400 500 600

B

N83D S292AS299A N83D+S292A

Nor

mal

ized

Abs

orba

nce

0

1

2

3

Wavelength (nm)

300 400 500 600

Killer whaleBovine

487 nm 499 nm

Acid bleach difference spectra

Acid

C

Abs

orba

nce

−0.02

−0.01

0

0.01

0.02

0.03

Wavelength (nm)300 400 500 600

Abs

orba

nce

−0.01

0

0.01

Wavelength (nm)

400 500 600

FIG. 1. Absorption spectra for wild-type killer whale and bovine control showing bleaching by (A) light and (B) acid, and absorption spectra for (C)mutant killer whale rhodopsins. The indicated �max values were estimated according to the curve-fitting methodology of Govardovskii et al. (2000).The red lines show the shifted spectra that result after sample exposure to white light (�max = 380 nm), or to 100 mM hydrochloric acid(�max = 440 nm). Dark-light and dark-acid difference spectra are shown to the right in panel (A) and (B), and as insets in panel (C).

Table 1. Spectral Tuning of Killer Whale and Bovine RhodopsinsMeasured In Vitro.

Species Mutant kmax 83, 292, 299

Bos taurus Wild-type 499 D A A

Orcinus orca N83D+S292A 501 D A S

Orcinus orca S292A 497 N A S

Orcinus orca N83D 489 D S S

Orcinus orca Wild-type 487 N S S

Orcinus orca S299A 485 N S A

NOTE.—�max values presented for this study are the mean of three dark spectrummeasurements. 83, 292, and 299 refer to the amino acid positions of the rhodopsinprotein sequence. The amino acid identities are indicated with standard abbrevia-tions (A, alanine; D, aspartic acid; N, asparagine; S, serine).

3

Spectral Tuning of Killer Whale Rhodopsin . doi:10.1093/molbev/msv217 MBE at U




ownloaded from









positive selection on a subset of sites (M2a vs. M1a, M8 vs.M7, M8 vs. M8a, P< 0.01; table 2 and supplementarytable S2, Supplementary Material online). We found sim-ilar results using both a species tree (fig. 3) of commonlyaccepted cetacean relationships (McGowen 2011; Gatesy

et al. 2013) as well as a rhodopsin gene tree (supplemen-tary fig. S2, Supplementary Material online). HYPHY’sPARRIS test (Scheffler et al. 2006), which resembles theM2a versus M1a test but with variation in synonymousrates (dS) incorporated, was also significant (P< 0.05,

FIG. 2. Homology model of killer whale rhodopsin. (A) Overall, bovine (gray) and killer whale (blue) rhodopsin were closely aligned. (B–E) Killer whaleresidues at spectral tuning sites 292 and 299 add hydroxyl groups near the chromophore (orange) and reduce side-chain distance to the Schiff base (SB).(F) Superposition of killer whale (solid) and bovine (transparent) rhodopsin showing participation of site 83 in a hydrogen-bond network (waters and H-bonds shown as blue spheres and dotted lines, respectively) that may influence chromophore isomerization (Palczewski et al. 2000; Okada et al. 2002).The different residues in killer whale (N83) and bovine (D83) appear to only minimally impact the configuration of this network.

4





ownloaded from








supplementary table S3, Supplementary Material online),indicating that this result is robust to variation in dS.

Positively Selected Sites and Spectral Tuning

Codon-based random sites models were also used to identifyspecific amino acid sites under positive selection. In total,we found 29 positively selected sites in cetacean rhodopsinas identified with a posterior probability of at least 80% ofbelonging to a positively selected site class under PAML’s M8model (8 sites), HYPHY’s REL (26 sites) or FUBAR (10 sites)models (table 3 and supplementary fig. S3, SupplementaryMaterial online). Three sites (16, 104, and 299) were identifiedby all three of the models, and 12 were identified by at leasttwo of the models (fig. 4 and table 3). Six of the M8 sites and allof the FUBAR sites were identified by at least one of the othertwo models, but the less stringent REL model identified 15sites that were not detected by the other two models. The twoM8-only sites (194 and 196) had low posterior probabilitiesand !< 1 under the HYPHY models. Generally, sites identi-fied as positively selected in both the HYPHY models but notM8 were nevertheless assigned to the positively selected siteclass in M8, only with posterior probabilities less than 80%.

Of the three spectral tuning sites investigated by site-directed mutagenesis in our in vitro experiments, site 299was identified as positively selected in all three models. Site292 was identified by M8 and REL, but not FUBAR (though !was still slightly 41). Site 83 had the weakest support forpositive selection, and was only identified by REL; thoughslightly elevated relative to the majority of other sites, ! wasestimated as less than 1 in both M8 and FUBAR. The levels ofsupport for these three sites are likely related to their differentlevels of variation on the cetacean species tree. The codon-based tests for positive selection that we have used are opti-mized to detect diversifying rather than directional selection,and so are most powerful when sites undergo frequent sub-stitutions, including convergences and reversals. Sites 292 and299 both show greater variation on the cetacean tree than 83.However, despite the varying support for positive selection atthese three sites, all were placed in the divergent site classunder the CmC (Clade model C) depth models (see below).

Clade Model Tests of Divergence

We used clade models to investigate the relationship betweendivergence and ecological variables. We hypothesized thatecological differences among cetaceans occupying differentaquatic ecosystems, and that forage at different depths,may be driving divergence in rhodopsin evolution. To testthese hypotheses, we implemented CmC, designing partitionsto reflect differences in habitat (oceanic, coastal, and fresh-water) and foraging depth (epipelagic, mesopelagic, andbathypelagic) (fig. 3). For the likelihood ratio tests (LRTs)among nested models, only the three- and four-partitiondepth models performed better than the simplest (null)model, M2a_rel, which suggests that foraging depth, ratherthan habitat or phylogeny, is the main factor driving diver-gence among cetacean rhodopsins. In the four-partitionmodel dN/dS was lowest for the mesopelagic clades (nearneutral at != 0.68), but increased above 1 for the epipelagicand bathypelagic foragers, the bathypelagic foragers havingthe greatest value (!= 2.54; table 4, “CmC: Depth 4-partition”). However, the three-partition model that groupedbathypelagic and epipelagic clades together was a significantlybetter fit than the four-partition model, as well as the two-partition model with only the bathypelagic+epipelagic cladein the foreground (table 4, “CmC: Depth 3-partition,” “CmC:Depth 2-partition”). Two- and three-partition models thatplaced epipelagic and bathypelagic clades in the foregroundseparately also did not show significant divergence (data notshown). In other words, although divergent selection signalswere strongest among near-surface foragers and extremedivers, mesopelagic foraging clades were also contributingto the significant divergence pattern. Evaluating the three-partition depth model against a null model where the bathy-pelagic+epipelagic divergent site class was constrained toequal one was also significant (table 4, “CmC: Depth3-partitionC”), confirming the presence of positive selectionwithin the bathypelagic and epipelagic foragers. The best fit-ting model overall (according to Akaike’s InformationCriterion, AIC) was the three-partition foraging depthmodel (table 4, “CmC: Depth 3-partition”), with the !value for bathypelagic+epipelagic clade being similar to that

Table 2. LRTs for Random-Sites Models (PAML) of the Cetacean RH1 Species Tree.

Model np ln L j Parametersa Null LRT df P

x0/p x1/q x2/xp

M0 49 �3,166.43 4.12 0.12

M1a 50 �3,052.82 4.29 0.02 (88.0) 1 (12.0)

M2a 52 �3,047.29 4.35 0.02 (87.9) 1 (11.6) 7.13 (0.5) M1a 11.1 2 0.004*

M3 53 �3,048.16 4.34 0.02 (87.5) 0.92 (11.9) 6.85 (0.6) M0 236.5 4 0.000*

M7 50 �3,054.33 4.27 0.05 0.33

M8a 51 �3,052.29 4.29 0.18 5.64 1.0 (10.4)

M8 52 �3,048.73 4.35 0.08 0.76 2.46 (2.9) M7 11.2 2 0.004*M8a 7.1 1 0.008*

NOTE.—np, number of parameters; ln L, ln likelihood; �, transition/transversion ratio; df, degrees of freedom.aFor models M0–M3, the ! values for each site class (!0–!2) are shown with their proportions in parentheses. For models M7–M8, p and q describe the shape of the betadistribution, and !p refers to the positively selected site class (with proportion in parentheses) for models M8 and M8a (where it is constrained to one). *Indicates statisticalsignificance (p< 0.05).

5





ownloaded from







estimated for the positively selected site class in the M8model.

Cetacean rhodopsin evolution did not appear to be signif-icantly driven by the evolutionary differences betweentoothed and baleen species, as there was no evidence fordivergent selection occurring between the two cetacean sub-orders, Odontoceti and Mysticeti using CmC (table 4).Branch-site models were also implemented, as they are in-creasingly being used to investigate the evolution of bothbranches and clades (e.g., Spady et al. 2005; Badouin et al.2013; Veilleux et al. 2013; Schott et al. 2014). However, thebranch-site model can only accommodate two partitions, sowe could not evaluate any of our higher partition configura-tions this way. Only the model with Odontoceti in the fore-ground turned out to fit significantly better than the nullmodel (P< 0.05; supplementary table S4, SupplementaryMaterial online), even though the equivalent clade modeltest was not significant. Of the branch-site models, themodel with all Cetacea in the foreground had the best fitaccording to AIC, and was the only branch-site model thatperformed better than its equivalent CmC model under this

criterion. For all but the Cetacea-foreground partition, thebranch-site models were also worse fits than the random-sites M2a-rel model, which assumes no partitions in thedata set (supplementary table S4, Supplementary Materialonline). Because these models do not allow background pos-itive selection (even though both Odontoceti and Mysticetiappear to have ! values greater than 1, with Odontocetibeing more pronounced), the poor fits relative to the nullare consistent. Overall the branch-site models seemed to havea weaker performance on our data set, and our clade modelresults suggest that this is because divergent patterns in !cannot be accommodated by its restrictions.

DiscussionBy combining mutation experiments and protein functionassays with molecular evolutionary analyses, we have foundevidence for functional adaptation in cetacean rhodopsin.The rhodopsin of the killer whale (Orcinus orca), a delphinid,was expressed in vitro for the first time, along with a series ofmutants with substitutions at the hypothesized spectraltuning sites, 83, 292, and 299. We found a maximal

Balaenoptera musculus

Berardius bairdii

Delphinus delphis

Balaenoptera acutorostrata scammoni

Delphinapterus leucas

Balaenoptera physalus

Balaenoptera acutorostrata acutorostrata

Ziphius cavirostris

Lipotes vexillifer

Pontoporia blainvillei

Globicephala melas

Mesoplodon bidens

Caperea marginata

Eubalaena glacialis

Hippopotamus amphibius

Neophocaena phocaenoides

Inia geoffrensis

Phocoena phocoena

Physeter macrocephalus

Bos taurus

Platanista minor

Tursiops truncatus

Phocaenoides dalli

Orcinus orcaOutgroupsEpipelagicMesopelagicBathypelagic

OutgroupsFreshwater

CoastalOceanic

M

O

C

FIG. 3. Tree topology used in codon-based likelihood models showing lineage partitions according to foraging depth (left) and habitat (right). Zonedefinitions: Epipelagic (0–200 m), mesopelagic (200–1,000 m), bathypelagic (41,000 m), oceanic (beyond continental shelves), coastal (within conti-nental shelves), freshwater (rivers, lakes, estuaries). Please see supplementary table S5, Supplementary Material online, for details regarding assignment oflineages to different groups. C, Cetacea; M, Mysticeti; O, Odontoceti. Note that the unrooted tree is required for model implementation.

6





ownloaded from









absorbance of 487 nm in the wild-type pigment, which wasconsiderably blue-shifted relative to a terrestrial mammal(bovine, 499 nm). Site-directed mutagenesis experiments inthe killer whale rhodopsin background showed that muta-tions at the three tuning sites were found to account for thespectral difference between the killer whale and bovine rho-dopsin. These results are consistent with previous mutagen-esis studies of these sites in a bovine background (Fasick andRobinson 1998), as well as with other wild-type cetaceanrhodopsins with equivalent substitutions (Bischoff et al.2012), although we did find variation in the magnitude ofspectral shifts. Furthermore, we found significant evidence forpositive selection in a data set of cetacean rhodopsins withup to 29 positively selected sites, including 83, 292, and299. We also found evidence to suggest that divergent selec-tion in the gene is driven, at least in part, by ecological differ-ences related to diving behavior. Here, we discuss our resultsin the context of both rhodopsin structure-function andcetacean ecology.

The Role of Sites 83, 292, and 299 in CetaceanRhodopsin Spectral Tuning

We expressed a killer whale visual pigment for the first time,and found the spectral tuning of its rhodopsin (487 nm) to besubstantially blue-shifted relative to bovine rhodopsin(499 nm). This is similar to other delphinids whose pigmentshave been expressed and assayed in vitro (488–489 nm; Fasicket al. 1998; Fasick and Robinson 2000). In the past, this differ-ence between terrestrial and marine mammal rhodopsin hasbeen attributed to substitutions at sites 83, 292, and 299. Notonly are these sites implicated in the spectral tuning of rho-dopsin in other aquatic vertebrates (e.g., Hunt et al. 1996,2001; Hope et al. 1997; Sugawara et al. 2005), but mutationsat these sites in bovine rhodopsin are also thought to mimicthe range of observed cetacean rhodopsin �max (Fasick andRobinson 1998, 2000). In addition, Meredith et al. (2013)found significant evidence for positive selection on sites 83and 292 (but not 299) in cetacean rhodopsin using branch-sites models. Residues at these sites have been used to predict

Table 3. x Values for Positively Selected Sites under Random-Sites Models from PAML (M8) and HYPHY (REL, FUBAR).

M8 REL FUBAR

Site x Post. Probability x Post. Probability x Post. Probability

7 1.063 0.437 2.055 0.962 1.238 0.533

16a 1.939 0.968 1.711 0.950 2.811 0.877

26 0.547 0.150 1.703 0.924 0.921 0.628

39 0.450 0.100 1.883 0.910 1.248 0.625

49 0.449 0.098 1.933 0.915 1.339 0.638

83 0.777 0.246 1.725 0.951 0.955 0.682

99b 1.631 0.760 2.190 0.996 4.335 0.877

104a 1.883 0.931 1.983 0.978 8.014 0.966

112 1.641 0.765 2.079 0.985 2.649 0.745

137 0.470 0.110 1.957 0.921 1.392 0.646

151 0.346 0.049 1.548 0.848 0.659 0.561

159 0.291 0.027 1.925 0.937 1.374 0.651

162 0.250 0.015 1.719 0.893 0.952 0.601

165b 1.463 0.659 2.174 0.994 3.344 0.844

194 1.826 0.891 0.620 0.595 0.789 0.375

195b 1.973 0.998 0.658 0.630 5.500 0.887

196 1.936 0.963 0.294 0.011 0.320 0.010

198 0.514 0.085 2.018 0.980 1.884 0.751

213b 1.076 0.423 2.018 0.981 3.327 0.869

216 1.153 0.474 2.028 0.981 2.098 0.766

259b 1.748 0.831 1.227 0.851 0.524 0.235

266b 1.303 0.564 2.132 0.990 2.779 0.816

282 1.198 0.498 1.728 0.952 1.398 0.772

290b 0.911 0.314 1.829 0.963 2.165 0.814

292b 1.863 0.910 1.322 0.883 1.037 0.498

299a 1.845 0.903 2.031 0.982 7.063 0.959

329 1.644 0.766 1.955 0.876 1.733 0.706

333b 0.885 0.297 2.031 0.982 2.611 0.830

335 0.552 0.108 2.022 0.980 1.685 0.728

NOTE.—Sites were identified as positively selected if they had a posterior probability greater than 80% of belonging to the positively selected site class.aSite was identified by all three models.bSite was identified by two models.

7





ownloaded from


the rhodopsin �max of different cetacean species (e.g., Fasicket al. 2011; Bischoff et al. 2012), though for many cetaceansthis has yet to be confirmed by in vitro expression experi-ments. Variation at site 292 has been associated with blueshifts in the rhodopsins of deep-dwelling fish that can rangefrom 11 to 14 nm depending on the species (Sugawara et al.2005). Site 299 has not been as well investigated in previousexperimental studies; however, given its increased distancefrom the chromophore relative to site 292, its lesser effecton �max in our study was not surprising. The killer whaleresidue, N83, has also long been associated with varying de-grees of blue-shifts in deep-dwelling teleost fish rhodopsins(Hunt et al. 1996, 2001; Hope et al. 1997; Sugawara et al. 2005),and in the nocturnal echidna, mutating N83D red-shifts �maxby as much as 6 nm (Bickelmann et al. 2012). The mutagenesisdata for site 83 in particular illustrate that predicting �maxbased on sequence alone can be a risky proposition, as thespectral effects of substitutions at a site can vary substantiallydepending on the genetic background in which those substi-tutions were made (Asenjo et al. 1994; Hauser et al. 2014).

In this study, we used site-directed mutagenesis to inves-tigate the contributions of sites 83, 292, and 299 to the blue-shifted killer whale �max relative to terrestrial mammals. Ourresults support the notion that although these sites areclearly the primary determinants of spectral tuning, theremay be other sites that interact with these to affect cetaceanrhodopsin �max. Though the directions of the �max shifts inour killer whale mutants were consistent, the magnitudes ofthe shifts were found to differ compared with the reversemutations in bovine rhodopsin (Fasick and Robinson 1998).The degree of shifts present in our mutant killer whale rho-dopsins also accounts for some, but not all of the spectralshifts found in other expressed wild-type cetacean rhodop-sins. For example, the North Atlantic right whale (Eubalaenaglacialis) has N83, A292, and S299, which according to ourmutagenesis results should result in a �max of approximately

497 nm, but Bischoff et al. (2012) measured it as 493 nm. Awide variety of sites have been shown to influence visualpigment �max in various vertebrate species (Hunt et al.2001, 2009; Bowmaker and Hunt 2006; Bowmaker 2008),and some of these clearly warrant further investigation incetaceans. In addition, spectral tuning sites (site 83 in partic-ular) have also been implicated in nonspectral functions suchas active state (Meta II) kinetics (e.g., Sugawara et al. 2010),but these functions, and their potential adaptive relevance,have yet to be investigated in cetaceans.

Positive Selection in Cetacean Rhodopsin

Our results demonstrated significant evidence for posi-tive selection in cetacean rhodopsins. Even though thereare many mammalian groups with specialized visual abil-ities, particularly for dim-light environments, evidencefor positive selection has not typically been found formammalian rhodopsin data sets, and overall substitu-tion rate ratios tend to be very low (e.g., M0 != 0.04in Zhao et al. [2009] vs. our data set M0 != 0.12). Onepossibility is that selection signatures in such data setsare too weak to be detected, especially if taxonomic cov-erage is wide. Rhodopsin selection signals are generallyweaker relative to other proteins due to the majority ofthe amino acid sequence being highly conserved. Inother vertebrate groups, population-level polymor-phism data have been crucial for detecting signaturesof selection among closely related species for highly con-served genes, including rhodopsin (Larmuseau et al.2010). Our random-sites models indicated that only asmall portion of sites (3%) were under positive selection.This is consistent with other positively selected verte-brate rhodopsins (e.g., Spady et al. 2005; Rennison et al.2012; Schott et al. 2014), but does not approach the highproportions and substitution rate ratios typically foundin rapidly diversifying genes, such as those that encode

FUBARRELM8

16*

99

104*

165

195

213

259

266

290 292

299*

33383

ω v

alue

0

1

2

3

4

5

6

7

8

9

Amino acid site

0 50 100 150 200 250 300 350

FIG. 4. ! values for each site of cetacean rhodopsin. Shaded regions indicate transmembrane helix domains. Sites identified as positively selected with�80% posterior probability in at least two of the random-sites models, M8 (PAML), REL and FUBAR (HYPHY), are labeled (exception site 83, identifiedby REL only, shown for spectral tuning context). Asterisks indicate that the site was identified by all three models.

8





ownloaded from


major histocompatibility complex (MHC) proteins(Swanson et al. 2001). More recent work has also re-vealed that mammalian rhodopsins have elevated syn-onymous substitution rates and experience positiveselection at synonymous sites (Du et al. 2014); nonsy-nonymous positive selection may be more difficult todetect as a consequence of this pattern. Importantsites may not always be detected by computational anal-yses of selection patterns, making experimentally de-rived protein structure-function data critical formolecular evolutionary studies. Our recovery of a signif-icant signal of positive selection highlights the suitabilityof cetaceans for studies of molecular adaptation, as theyare a group that is both ecologically diverse and rela-tively closely related (McGowen et al. 2014).

Although a variety of other genes show evidence of posi-tive selection in cetaceans (reviewed in McGowen et al. 2014),this has not typically been the case for the visual pigments,despite experimental evidence for functionally important

variation. In particular, tests for positive selection usingrandom-sites models have not been significant, though evi-dence was found for positive selection using branch-sitemodels (Meredith et al. 2013). Our data set indicated thathomoplasies (e.g., convergences and reversals) were frequentenough in the cetacean RH1 gene to result in an inaccuratetree topology, an idea that was supported by recovery of amore accurate topology when nonsynonymous nucleotidepositions were removed from the alignment (supplementaryfig. S2, Supplementary Material online). Interestingly, conver-gent evolution appears to have led to recovery of inaccuratemammalian relationships in the past, not only in rhodopsin(Zhao et al. 2009; Shen et al. 2010) but also in other positivelyselected genes such as prestin (SLC26A5), which is linked tohigh-frequency hearing (Liu, Cotton, et al. 2010; Elgoyhen andFranchini 2011). Convergence signals are so strong in mam-malian prestin that the gene tree actually has a topology thatplaces echolocating bats and cetaceans as sister lineages(Jones 2010; Li et al. 2010; Liu, Rossiter, et al. 2010). The

Table 4. Results from Clade Model (CmC—PAML) Tests for Divergence Partitioned by Phylogeny, Foraging depth, and Habitat.

Model and Partitiona np ln L j Parametersb �AICc Null LRT df P

x0 x1 x2/xd

M2a_rel 52 �3,047.29 4.35 0.02 (87.9) 1 (11.6) 7.13 (0.5) 4.34

CmC: Phylo (4-partition) 55 �3,047.15 4.31 0.02 (88.2) 1 (3.7) 0.26 (8.1) 10.06 M2a_rel 0.28 3 0.964Root: 1.08 2-partition 0.26 2 0.878Myst: 1.46 3-partition 0.16 1 0.689Odon: 1.57

CmC: Phylo (3-partition) 54 �3,047.23 4.31 0.02 (88.1) 1 (3.8) 0.24 (8.1) 8.22 M2a_rel 0.12 2 0.942Myst+Root: 1.33 2-partition 0.10 1 0.752Odon: 1.56

CmC: Phylo (2-partition) 53 �3,047.28 4.35 0.02 (87.9) 1 (11.2) 5.09 (0.9) 6.32 M2a_rel 0.02 1 0.888Odon: 7.85

CmC: Habitat (4-partition) 55 �3,045.78 4.31 0.02 (86.5) 1 (6.8) 0.0 (6.6) 7.32 M2a_rel 3.02 3 0.389Ocean: 0.71 2-partition 2.88 2 0.237Coast: 1.23 3-partition 2.88 1 0.090Fresh: 2.48

CmC: Habitat (3-partition) 54 �3,047.22 4.31 0.02 (86.7) 1 (6.7) 0.0 (6.4) 8.20 M2a_rel 0.14 2 0.932Salt: 0.91 2-partition 0.00 1 1.000Fresh: 2.43

CmC: Habitat (2-partition) 53 �3,047.22 4.33 0.02 (87.3) 1 (11.2) 0.0 (1.5) 6.20 M2a_rel 0.14 1 0.708Fresh: 2.41

CmC: Depth (4-partition) 55 �3,042.96 4.33 0.02 (87.7) 1 (5.0) 0.09 (7.1) 1.68 M2a_rel 8.66 3 0.034*Epi: 1.91 2-partition 6.84 2 0.033*Meso: 0.68 3-partition 0.32 1 0.572Bath: 2.54

CmC: Depth (3-partition) 54 �3,043.12 4.33 0.02 (87.8) 1 (5.1) 0.09 (7.2) 0.00 M2a_rel 8.34 2 0.015*Meso: 0.69 2-partition 6.52 1 0.011*Epi+Bath: 2.12 3-partitionC 4.54 1 0.033*

CmC: Depth (3-partitionC) 53 �3,045.39 4.33 0.01 (85.4) 1 (6.7) 0.0 (7.9) 2.54 M2a_rel 3.80 1 0.051Meso: 0.47Epi+Bath: 1.00

CmC: Depth (2-partition) 53 �3,046.38 4.34 0.02 (87.9) 1 (5.7) 0.40 (6.4) 4.52 M2a_rel 1.82 1 0.177Epi+Bath: 2.12

NOTE.— np, number of parameters; ln L, ln likelihood; �, transition/transversion ratio; df, degrees of freedom; Cet, Cetacea; Odon, Odontoceti; Myst, Mysticeti; Root, root branchof cetacean; Ocean, oceanic; Coast, coastal; Fresh, freshwater/estuarine; Meso, mesopelagic; Epi, epipelagic; Bath, bathypelagic.aPartitions for Habitat and Depth are explained in figure 3, and for 2- and 3-partition categories, the model with the highest ln-likelihood is shown.b! values for each site class (!0–!2) are shown with the proportion of each in parentheses. !d refers to the divergent site class in the CmC models, which has a separate valuefor each partition: The first value is for the background, followed by the foreground clade(s).cMinimum overall AIC (foraging depth 3-partition of CmC; 6,194.24) was used for all comparisons.*Indicates statistical significance (p< 0.05).

9





ownloaded from





rhodopsin gene is also known for having sites with convergentpatterns in other vertebrate data sets where it is positivelyselected (Larmuseau et al. 2011; Schott et al. 2014).Studies such as these suggest that, in certain cases, speciestrees may be more appropriate than gene trees when direc-tional evolutionary pressure may skew phylogeneticreconstructions.

The Role of Rhodopsin in Cetacean Visual Ecology

In addition to statistical and experimental evidence forpositive selection and functional adaptation in cetacean rho-dopsin, we also found evidence for higher dN/dS in both near-surface and extremely deep foragers. Absorption of light bywater increasingly narrows the light spectrum to predomi-nantly blue wavelengths as depth increases, and in clear oceanattenuates light intensity such that deeper than approxi-mately 200 m, the boundary of the epipelagic zone, photo-synthesis is no longer supported (Wozniak and Dera 2007;Warrant and Johnsen 2013). As such, the general expectationis that organisms that live or are active in the deep-sea willhave blue-shifted visual sensitivity to capitalize on the mini-mal amounts of available light (Lythgoe and Dartnall 1970;Hunt et al. 2001; Griebel and Peichl 2003). Codon-based clademodels are increasingly being used to test hypotheses of eco-logical divergence (Schott et al. 2014; Van Nynatten et al.2015), and in our case they were a useful exploratory toolfor determining whether there was statistical evidence forrhodopsin divergence based on foraging depth. Significantevidence for positive selection in both the bathypelagic andepipelagic clades suggested that divergent selection signalswere strongest in these lineages, likely reflecting commondescent from primarily mesopelagic foraging ancestors (seesupplementary table S5, Supplementary Material online).Nevertheless, the models that showed significant evidencefor divergence also included a mesopelagic partition withslightly elevated dN/dS, which suggests that there is function-ally important variation distinguishing this foraging profile aswell. Even though the coarse partitions we used do not cap-ture the more nuanced variation in cetacean foragingbehavior, our results provide statistical evidence for an asso-ciation between rhodopsin evolution and foraging depth incetaceans that was previously only speculative. Furthermore,although prior studies have emphasized the bathypelagic for-agers (e.g., Fasick and Robinson 2000), our results suggest thatthe same subclass of sites is also positively selected in near-surface foragers. This makes sense under a spectral tuningregime where substitutions at a few sites can result in awide range of spectral sensitivities. In the future, with a greatersampling of cetacean species, this association could be furthertested and refined.

In addition to foraging, we also investigated divergencealong habitat partitions (oceanic, coastal, and freshwater).The spectrum and intensity of light can be affected by thescattering tendencies of dissolved particulate matter, which ismore abundant in coastal and freshwater systems due toterrestrial runoff (e.g., the red-shifted light environment ofsome river systems; Costa et al. 2013). However, we did not

find significant evidence for divergence with respect to thisaspect of ecology, though this model was still a better fit thanthe model partitioned by phylogeny. If habitat variables areinfluencing cetacean rhodopsin evolution, it is possible we didnot have enough power to detect this in our data set. Inparticular, our sampling of freshwater and estuarine cetaceanswas low and did not include any of the independent invasionsmade by the Delphinidae (e.g., Sotalia, Orcaella, Sousa). Onthe other hand, this lack of significance also suggests that thesignificant contribution of the epipelagic clades to the forag-ing depth models was also not due solely to the freshwaterspecies. Given that freshwater invasions have been implicatedin the spectral tuning of rhodopsin in several cichlid fish lin-eages (Spady et al. 2005; Terai et al. 2006; Miyagi et al. 2012;Schott et al. 2014) and in anchovies (Van Nynatten et al.2015), further investigations with a more complete data setof freshwater cetacean rhodopsins could be fruitful in thefuture.

The ways in which light can vary in underwater environ-ments have had a profound impact on the evolution of eyesin aquatic organisms (Warrant and Locket 2004), andcetaceans are no exception. Like fish, they possess a sphericallens (Mass and Supin 2007), and like many carnivorous mam-mals that forage in dim-light conditions, the retina isdominated by rod photoreceptors (Peichl 2005) and is sup-ported by a reflective tapetum lucidum layer (Mass and Supin2007). Among the visual pigments, blue-shifts in �maxare present not only in rhodopsin but also in the long-wavesensitive cone opsin as well (Fasick and Robinson 1998;Fasick et al. 1998; Newman and Robinson 2005).Oddly, marine mammals (cetaceans and pinnipeds) shareloss of function in the short-wave sensitive cone opsin withother dim-light inhabiting (nocturnal, fossorial) mammals(Jacobs 2013), which seems contradictory to the hypothesisof optimizing sensitivity toward downwelling light. Althoughother hypotheses have been put forward to explain the loss ofthis opsin (Peichl 2005), ultimately it remains a mystery;one that has been compounded by the recent discovery oflong-wave cone opsin losses in some cetacean lineages(Meredith et al. 2013). Addressing these issues in cetaceanvisual pigment evolution will certainly require a morethorough understanding of pigment function even beyondspectral tuning, in the context of the varied ecologies inwhich cetaceans have evolved. Some have argued thatthe field of molecular evolution in general requires an in-creased emphasis on experimental studies that test poten-tial targets of natural selection in genomic contexts byevaluating the phenotypic consequences of individualmutations (reviewed in Barrett and Hoekstra 2011). Wehave presented here the first results of mutation experimentson a cetacean rhodopsin in one of few studies of adap-tive molecular evolution in cetaceans to combine both bio-informatic and experimental approaches. All together, ourresults both computationally and experimentally supportadaptive evolution in cetacean rhodopsin to dim-lightvision, and suggest mechanistic and ecological selection pres-sures that can form the basis of hypotheses for futureinvestigations.

10





ownloaded from




Materials and Methods

Killer Whale RH1 Preparation and Site-DirectedMutagenesis

The wild-type killer whale rhodopsin gene (RH1) coding se-quence, along with a C-terminal nine-amino acid epitope tagfor the 1D4 antibody, was synthesized (GeneArt, Invitrogen)and inserted into the p1D4-hrGFP II expression vector(Morrow and Chang 2010). Site-directed mutagenesis(QuickChange II, Agilent) was used to generate mutantsN83D, S292A, S299A, as well as a double mutant N83D/S292A in the killer whale wild-type RH1. All mutants wereconfirmed by double-stranded sequencing.

Killer Whale Rhodopsin Expression and SpectralTuning

Killer whale rhodopsin and mutants were transientlytransfected, harvested, and purified alongside a bovine rho-dopsin control, as reported previously (Morrow and Chang2010). Briefly, vectors containing rhodopsin were transientlytransfected into HEK293T cells (Lipofectamine 2000;Invitrogen), harvested, regenerated with 11-cis-retinal, solubi-lized in 1% dodecylmaltoside, and immunoaffinity purifiedusing the 1D4 monoclonal antibody. The UV-visible absorp-tion spectra of purified rhodopsin samples were mea-sured using a Cary 4000 double-beam spectrophotometer(Varian) at 25 �C in the dark, and after 60 s of white lightbleaching. Difference spectra were calculated by subtractinglight spectra from dark spectra. �max was estimated byfitting a template curve (Govardovskii et al. 2000) to thenormalized dark spectrum. The wild-type killer whale rhodop-sin was also exposed to hydrochloric acid (HCl; 100 mM)and hydroxylamine (NH2OH; 50 mM) to confirm thepresence of a properly functioning Schiff base link and pro-tection of the chromophore from hydrolyzing solvent,respectively.

Homology Modeling of Wild-Type and Mutant KillerWhale Rhodopsin

The killer whale rhodopsin 3D structure was inferred withhomology modeling in the Modeller package (Sali andBlundell 1993). Wild-type bovine rhodopsin (PDB code:IU19, Okada et al. 2004) was used as a modeling template.For each structure, 100 models were generated and ranked bythe Modeller objective function (optimized ten times/model),and the run with the lowest DOPE score (Shen and Sali 2006)was selected for visualization in MacPyMol (DelanoScientific). The bovine template and killer whale rhodopsinhad comparable total energies (comparable z-scores inProSA-web; Wiederstein and Sippl 2007), and stereochemicalconformations of bonds among amino acid residues in themodels had high probabilities (assessed in ProCheck;Laskowski et al. 1993), both indicators that the models wereof high quality.

Rhodopsin Sequence Alignment and PhylogeneticAnalyses

Full rhodopsin coding sequences (1,044 bp) for 23 cetaceanspecies were obtained from publicly available databases (sup-plementary table S1, Supplementary Material online), whichcovered most cetacean families. Coding sequences were alsoobtained from 22 other mammal species to represent majormammalian outgroup taxa during gene tree building (supple-mentary fig. S2, Supplementary Material online). Codons werealigned using PRANK as implemented on the webPRANKserver (L€oytynoja and Goldman 2010) with default settings.Our alignment of a total 45 sequences was used to generatetwo rhodopsin gene trees, one estimated using a maximum-likelihood method, and the other with a Bayesian method.The maximum-likelihood tree was constructed in PhyML 3.0(Guindon et al. 2010) under the GTR+I+G model with a BioNJstarting tree, the best of either NNI or SPR tree improvement,and aBayes branch support. The GTR+I+G model wasselected according to AIC values after running the alignmentthrough MrModelTest (Nylander 2004) in PAUP 4.0(Swofford 2002). The Bayesian analysis was run in MrBayes3.0 (Huelsenbeck and Ronquist 2001) using a reversible jumpMarkov chain Monte Carlo algorithm with a gamma rateparameter (nst = mixed, rates = gamma, four chains) for tworuns each of 2 million generations (25% burn-in, samplingevery 100 generations). Convergence was confirmed bychecking output diagnostics (standard deviation of splitfrequencies approached zero, scattered log-likelihoodplot).

Though the resulting gene trees were highly similar,there were several major incongruences with acceptedspecies relationships at deep nodes (supplementary fig.S2, Supplementary Material online) suggesting that con-vergence may have affected the gene tree, a finding thathas been previously reported for mammalian rhodopsins(Zhao et al. 2009; Shen et al. 2010). As with previousstudies (Shen et al. 2010), removing nonsynonymousnucleotide positions from the alignment resulted in agene tree that was much closer to accepted species re-lationships (supplementary fig. S2, SupplementaryMaterial online). However, in order to ensure that theslight topological differences did not significantly affectour results, random-sites analyses were performed onboth this gene tree and the accepted species tree.Before running selection analyses in PAML, outgroupswere pruned to include only the two nearest cetaceanrelatives, Bos taurus and Hippopotamus amphibius, sothat the tree could be unrooted (as required forcodeml algorithms) while maintaining the basal topol-ogy of Cetacea (fig. 3).

Molecular Evolutionary Analyses

Patterns of selection in the data set were investigated bysubmitting the alignment (supplementary fig. S3,Supplementary Material online) and pruned trees (fig. 3and supplementary fig. S2B, Supplementary Material online)

11





ownloaded from


















to the codeml program in the PAML 4.0 software package(Yang 2007). LRTs with a �2 distribution were conducted onnested random-sites models to determine evidence for vari-ation in ! (M3 vs. M0), and for positive selection (M2a vs.M1a, M8 vs. M7, and M8 vs. M8a). All models were run aminimum of three times with varying initial starting values for� (transition to transversion ratio) and! to ensure parameterestimates did not represent local optima in the samplingspace. For positive selection models, M2a and M8, PAMLincorporates a Bayes’ Empirical Bayes (BEB) analysis to identifyspecific sites (codons) under positive selection (Yang et al.2005).

Because PAML estimates dN/dS as a single variable, varia-tion in this parameter is usually assumed to be due to vari-ation in dN. As such, we also used models that relax thisassumption, in which dN and dS are estimated as separateparameters (REL, FUBAR, and PARRIS models in the HYPHYpackage; Pond and Frost 2005; Ben Murrell et al. 2013) asimplemented on the Datamonkey webserver (Delport et al.2010). The PARRIS model performs an LRT of positive selec-tion in a manner similar to the M1a versus M2a comparison,using three dS and two ! classes (!= 0 and != 1) in the nullmodel, and an additional freely varying! class in the alternatemodel (Scheffler et al. 2006). The REL and FUBAR models onlyestimate site-by-site! values (similar to PAML’s BEB analysis),and do not implement LRTs of selection patterns betweennested models. FUBAR assigns sites to a much greaternumber of site classes than either REL or the PAML models,arguably providing more accurate site-by-site ! estimates(Murrell et al. 2013). Under the models, M8, REL, andFUBAR, sites were identified as positively selected if theyhad a posterior probability�80% of belonging to a positivelyselected site class.

We used PAML’s CmC (Bielawski and Yang 2004) andbranch-site models (Zhang et al. 2005) to statistically testfor divergence along specific branches of the species treeused in the random-sites analyses. These models allow ! tovary not only among sites but also among foreground andbackground clades specified by the user. Under CmC, twoclasses of sites are allowed to evolve conservatively(0<!< 1) and neutrally (!= 1) across the whole tree, buta third site class is allowed to evolve differently among two ormore clade partitions. Significant divergence is indicated withan LRT comparing CmC with the null model, M2a_rel, whichdoes not allow ! to diverge in the third site class (Weadickand Chang 2012). We designed tests for CmC based on threedivergence hypotheses, each with four clade partitions:Phylogeny (Odontoceti, Mysticeti, root Cetacea, outgroups),foraging depth (bathypelagic, mesopelagic, epipelagic, out-groups), and habitat (freshwater, coastal, oceanic, outgroups)(fig. 3). Assignment of clades/nodes to habitat and depthclasses was based on a priori knowledge of cetacean ecologicalvariables for both living and extinct taxa (supplementary tableS5, Supplementary Material online). For each hypothesis, thefour-partition models were compared with nested three- andtwo-partition models using LRTs (the best fitting model ineach partition category is shown in table 4). The models fromeach hypothesis were compared using AIC values to

determine whether ecological variables had a greater influ-ence on rhodopsin divergence than species relationships. Thebranch-site model, while similar to CmC, only allows onedivision into foreground and background lineages, so wecould only test the two-partition configurations from eachCmC hypothesis. In addition to the conservative and neutralsite classes, there are two additional site classes where theforeground is positively selected and the background is eitherconserved or neutral. Of note is that the branch-site model,unlike CmC, cannot accommodate positive selection in thebackground lineages. Like the random-sites models, all modelswere run at least three times varying initial � and ! values toensure local optima were avoided in the sampling space.

Supplementary MaterialSupplementary figures S1–S3 and tables S1–S5 are available atMolecular Biology and Evolution online (http://www.mbe.oxfordjournals.org/).

Acknowledgments

This work was supported by a Natural Sciences andEngineering Research Council (NSERC) Discovery Grant toB.S.W.C., an NSERC Postgraduate Scholarship to S.Z.D., anda Human Frontier Science Program Grant to B.S.W.C. The 11-cis-retinal was generously provided by Rosalie Crouch(Medical University of South Carolina).

ReferencesAnisimova M, Kosiol C. 2009. Investigating protein-coding sequence

evolution with probabilistic codon substitution models. Mol BiolEvol. 26:255–271.

Asenjo AB, Rim J, Oprian DD. 1994. Molecular determinants of humanred/green color discrimination. Neuron 12:1131–1138.

Badouin H, Belkhir K, Gregson E, Galindo J, Sundstr€om L, Martin SJ,Butlin RK, Smadja CM. 2013. Transcriptome characterisation of theant Formica exsecta with new insights into the evolution of desatur-ase genes in social Hymenoptera. PLoS One 8:e68200.

Baird RW. 2000. The killer whale: foraging specializations and grouphunting. In: Mann J, Connor RC, Tyack PL, editors. Cetacean socie-ties: field studies of dolphins and whales. Chicago (IL): University ofChicago Press. p. 127–153.

Barrett RDH, Hoekstra HE. 2011. Molecular spandrels: tests of adapta-tion at the genetic level. Nat Rev Genet. 12:767–780.

Bickelmann C, Morrow JM, M€uller J, Chang BSW. 2012. Functional char-acterization of the rod visual pigment of the echidna (Tachyglossusaculeatus), a basal mammal. Vis Neurosci. 29:211–217.

Bielawski JP, Yang Z. 2004. A maximum likelihood method for detectingfunctional divergence at individual codon sites, with application togene family evolution. J Mol Evol. 59:121–132.

Bischoff NN, Nickle BB, Cronin TWT, Velasquez SS, Fasick JIJ. 2012. Deep-sea and pelagic rod visual pigments identified in the mysticetewhales. Vis Neurosci. 29:95–103.

Bowmaker JK. 2008. Evolution of vertebrate visual pigments. Vision Res.48:2022–2041.

Bowmaker JK, Hunt DM. 2006. Evolution of vertebrate visual pigments.Curr Biol. 16:R484–R489.

Cassens I, Vicario S, Waddell VG, Balchowsky H, Van Belle D, Ding W,Fan C, Mohan RS, Sim~oes-Lopes PC, Bastida R, et al. 2000.Independent adaptation to riverine habitats allowed survival of an-cient cetacean lineages. Proc Natl Acad Sci U S A. 97:11343–11347.

Costa MPF, Novo EMLM, Telmer KH. 2013. Spatial and temporal vari-ability of light attenuation in large rivers of the Amazon.Hydrobiologia 702:171–190.

12





ownloaded from






http://www.mbe.oxfordjournals.org/

http://www.mbe.oxfordjournals.org/


Delport W, Poon AFY, Frost SDW, Pond SLK. 2010. Datamonkey 2010: asuite of phylogenetic analysis tools for evolutionary biology.J Gerontol. 26:2455–2457.

Du J, Dungan SZ, Sabouhanian A, Chang BS. 2014. Selection on synon-ymous codons in mammalian rhodopsins: a possible role in opti-mizing translational processes. BMC Evol Biol. 14:96.

Elgoyhen AB, Franchini LF. 2011. Prestin and the cholinergic receptor ofhair cells: positively-selected proteins in mammals. Hear Res.273:100–108.

Fasick JI, Bischoff N, Brennan S, Velasquez S, Andrade G. 2011. Estimatedabsorbance spectra of the visual pigments of the North Atlanticright whale (Eubalaena glacialis). Mar Mamm Sci. 27:E321–E331.

Fasick JI, Cronin TW, Hunt DM. 1998. The visual pigments of the bot-tlenose dolphin (Tursiops truncatus). Vis Neurosci. 15:643–651.

Fasick JI, Robinson PR. 1998. Mechanism of spectral tuning in the dol-phin visual pigments. Biochemistry 37:433–438.

Fasick JI, Robinson PR. 2000. Spectral-tuning mechanisms of marinemammal rhodopsins and correlations with foraging depth. VisNeurosci. 17:781–788.

Gatesy J, Geisler JH, Chang J, Buell C, Berta A, Meredith RW, Springer MS,McGowen MR. 2013. A phylogenetic blueprint for a modern whale.Mol Phylogenet Evol. 66:479–506.

Govardovskii VI, Fyhrquist N, Reuter T, Kuzmin DG, Donner K. 2000. Insearch of the visual pigment template. Vis Neurosci. 17:509–528.

Griebel U, Peichl L. 2003. Colour vision in aquatic mammals—facts andopen questions. Aquat Mamm. 29.1:18–30.

Guindon S, Dufayard J-F, Lefort V, Anisimova M, Hordijk W, Gascuel O.2010. New algorithms and methods to estimate maximum-likeli-hood phylogenies: assessing the performance of PhyML 3.0. Syst Biol.59:307–321.

Hamilton H, Caballero S, Collins AG, Brownell RL. 2001. Evolution ofriver dolphins. Proc R Soc Lond B Biol Sci. 268:549–556.

Hauser FE, van Hazel I, Chang BSW. 2014. Spectral tuning in vertebrateshort wavelength-sensitive 1 (SWS1) visual pigments: can wave-length sensitivity be inferred from sequence data? J Exp Zool B.322:529–539.

Hope AJ, Partridge JC, Dulai KS, Hunt DM. 1997. Mechanisms of wave-length tuning in the rod opsins of deep-sea fishes. Proc R Soc Lond BBiol Sci. 264:155–163.

Huelsenbeck JP, Ronquist F. 2001. MRBAYES: Bayesian inference of phy-logenetic trees. Bioinformatics 17(8):754–755.

Hunt DM, Carvalho LS, Cowing JA, Davies WL. 2009. Evolution andspectral tuning of visual pigments in birds and mammals. PhilosTrans R Soc Lond B Biol Sci. 364:2941–2955.

Hunt DM, Dulai KS, Partridge JC, Cottrill P, Bowmaker JK. 2001. Themolecular basis for spectral tuning of rod visual pigments in deep-sea fish. J Exp Biol. 204:3333–3344.

Hunt DM, Fitzgibbon J, Slobodyanyuk SJ, Bowmakers JK. 1996.Spectral tuning and molecular evolution of rod visual pigmentsin the species flock of cottoid fish in Lake Baikal. Vision Res.36:1217–1224.

Jacobs GH. 2009. Evolution of colour vision in mammals. Philos Trans RSoc Lond B Biol Sci. 364:2957–2967.

Jacobs GH. 2013. Losses of functional opsin genes, short-wavelengthcone photopigments, and color vision—a significant trend in theevolution of mammalian vision. Vis Neurosci. 30:39–53.

Jones G. 2010. Molecular evolution: gene convergence in echolocatingmammals. Curr Biol. 20:R62–R64.

Kawamura S, Yokoyama S. 1998. Functional characterization of visualand nonvisual pigments of American chameleon (Anolis carolinen-sis). Vision Res. 38:37–44.

Kito Y, Suzuki T, Azuma M, Sekoguti Y. 1968. Absorption spectrum ofrhodopsin denatured with acid. Nature 218:955–957.

Lamb TD, Pugh EN. 2004. Dark adaptation and the retinoid cycle ofvision. Prog Retin Eye Res. 23:307–380.

Larmuseau MHD, Huyse T, Vancampenhout K, Van Houdt JKJ,Volckaert FAM. 2010. High molecular diversity in the rhodopsingene in closely related goby fishes: A role for visual pigments inadaptive speciation? Mol Phyl Evol. 55(2):689–698.

Larmuseau MHD, Vanhove MPM, Huyse T, Volckaert FAM, Decorte R..2011. Signature of selection on the rhodopsin gene in the marineradiation of American seven-spined gobies (Gobidae,Gobiosomatini). J Evol Biol. 24(7):1618–1625.

Laskowski RA, Macarthur MW, Moss DS, Thorton JM. 1993. Procheck—a program to check the stereochemical quality of protein structures.J Appl Crystallogr. 26:283–291.

Li Y, Liu Z, Shi P, Zhang J. 2010. The hearing gene prestin unites echo-locating bats and whales. Curr Biol. 20:R55–R56.

Lindblad-Toh K, Garber M, Zuk O, Lin MF, Parker BJ, Washietl S,Kheradpour P, Ernst J, Jordan G, Mauceli E, et al. 2011. A high-resolution map of human evolutionary constraint using 29 mam-mals. Nature 478:476–482.

Liu Y, Cotton JA, Shen B, Han X, Rossiter SJ, Zhang S. 2010. Convergentsequence evolution between echolocating bats and dolphins. CurrBiol. 20:R53–R54.

Liu Y, Rossiter SJ, Han X, Cotton JA, Zhang S. 2010. Cetaceans on amolecular fast track to ultrasonic hearing. Curr Biol. 20:1834–1839.

L€oytynoja A, Goldman N. 2010. webPRANK: a phylogeny-aware multiplesequence aligner with interactive alignment browser. BMCBioinformatics 11:579.

Lythgoe JN, Dartnall HJ. 1970. A “deep sea rhodopsin” in a mammal.Nature 227:955–956.

Mass AM, Supin AY. 2007. Adaptive features of aquatic mammals’ eye.Anat Rec (Hoboken). 290:701–715.

McFarland WN. 1971. Cetacean visual pigments. Vision Res.11:1065–1076.

McGowen MR, Gatesy J, Wildman DE. 2014. Molecular evolution tracksmacroevolutionary transitions in Cetacea. Trends Ecol Evol.29:336–346.

McGowen MR. 2011. Toward the resolution of an explosive radiation—a multilocus phylogeny of oceanic dolphins (Delphinidae). MolPhylogenet Evol. 60:345–357.

Meredith RW, Gatesy J, Emerling CA, York VM, Springer MS. 2013. Rodmonochromacy and the coevolution of cetacean retinal opsins.PLoS Genet. 9:e1003432.

Miyagi R, Terai Y, Aibara M, Sugawara T, Imai H, Tachida H, Mzighani SI,Okitsu T, Wada A, Okada N. 2012. Correlation between nuptialcolors and visual sensitivities tuned by opsins leads to speciesrichness in sympatric Lake Victoria cichlid fishes. Mol Biol Evol.29:3281–3296.

Morrow JM, Chang BSW. 2010. The p1D4-hrGFP II expression vector: atool for expressing and purifying visual pigments and other G pro-tein-coupled receptors. Plasmid 64:162–169.

Moura AE, van Rensburg CJ, Pilot M, Tehrani A, Best PB, ThorntonM, Ploen S, de Bruyn PJN, Worley KC, Gibbs RA, et al. 2014.Killer whale nuclear genome and mtDNA reveal widespreadpopulation bottleneck during the last glacial maximum. MolBiol Evol. 31:1121–1131.

Murrell B, Moola S, Mabona A, Weighill T, Sheward D, Pond SLK,Scheffler K. 2013. FUBAR: a fast, unconstrained Bayesian approxi-mation for inferring selection. Mol Biol Evol. 30:1196–1205.

Newman LA, Robinson PR. 2005. Cone visual pigments of aquatic mam-mals. Vis Neurosci. 22:873–879.

Nilsson D-E. 2013. Eye evolution and its functional basis. Vis Neurosci.30:5–20.

Nylander JAA. 2004. MrModeltest v2. Uppsala (Sweden): UppsalaUniversity.

Okada T, Fujiyoshi Y, Silow M, Navarro J, Landau EM, Shichida Y.2002. Functional role of internal water molecules in rhodopsinrevealed by x-ray crystallography. Proc Natl Acad Sci U S A.99:5982–5987.

Okada T, Sugihara M, Bondar A-N, Elstner M, Entel P, Buss V. 2004. Theretinal conformation and its environment in rhodopsin in light of anew 2.2A crystal structure. J Mol Biol. 342:571–583.

Palczewski KK, Kumasaka TT, Hori TT, Behnke CAC, Motoshima HH,Fox BAB, Le Trong II, Teller DCD, Okada TT, Stenkamp RER, et al.2000. Crystal structure of rhodopsin: a G protein-coupled receptor.Science 289:739–745.

13





ownloaded from


Peichl L. 2005. Diversity of mammalian photoreceptor properties: adap-tations to habitat and lifestyle? Anat Rec 287A:1001–1012.

Pond SLK, Frost SDW. 2005. Not so different after all: a comparison ofmethods for detecting amino acid sites under selection. Mol BiolEvol. 22:1208–1222.

Pond SLK, Frost SDW, Muse SV. 2005. HyPhy: hypothesis testing usingphylogenies. Bioinformatics 21:676–679.

Rennison DJ, Owens GL, Taylor JS. 2012. Opsin gene duplication anddivergence in ray-finned fish. Mol Phylogenet Evol. 62:986–1008.

Sakmar TPT, Franke RRR, Khorana HGH. 1989. Glutamic acid-113 servesas the retinylidene Schiff base counterion in bovine rhodopsin. ProcNatl Acad Sci U S A. 86:8309–8313.

Sali A, Blundell TL. 1993. Comparative protein modeling by satisfactionof spatial restraints. J Mol Biol. 234:779–815.

Scheffler K, Martin DP, Seoighe C. 2006. Robust inference of positiveselection from recombining coding sequences. Bioinformatics22:2493–2499.

Schott RK, Refvik SP, Hauser FE, L�opez-Fern�andez H, Chang BSW. 2014.Divergent positive selection in rhodopsin from lake and riverinecichlid fishes. Mol Biol Evol. 31:1149–1165.

Shen M-Y, Sali A. 2006. Statistical potential for assessment and predic-tion of protein structures. Protein Sci. 15:2507–2524.

Shen Y-Y, Liu J, Irwin DM, Zhang Y-P. 2010. Parallel and convergentevolution of the dim-light vision gene RH1 in bats (Order:Chiroptera). PLoS One 5:e8838.

Spady TC, Seehausen O, Loew ER, Jordan RC, Kocher TD, Carleton KL.2005. Adaptive molecular evolution in the opsin genes of rapidlyspeciating cichlid species. Mol Biol Evol. 22:1412–1422.

Sugawara T, Imai H, Nikaido M, Imamoto Y, Okada N. 2010. Vertebraterhodopsin adaptation to dim light via rapid meta-II intermediateformation. Mol Biol Evol. 27:506–519.

Sugawara T, Terai Y, Imai H, Turner GF, Koblmuller S, Sturmbauer C,Shichida Y, Okada N. 2005. Parallelism of amino acid changes at theRH1 affecting spectral sensitivity among deep-water cichlids fromLakes Tanganyika and Malawi. Proc Natl Acad Sci U S A. 102:5448–5453.

Sun Y-B, Zhou W-P, Liu H-Q, Irwin DM, Shen Y-Y, Zhang Y-P. 2013.Genome-wide scans for candidate genes involved in the aquaticadaptation of dolphins. Genome Biol Evol. 5:130–139.

Swanson WJ, Yang Z, Wolfner MF, Aquadro CF. 2001. Positive Darwinianselection drives the evolution of several female reproductive pro-teins in mammals. Proc Natl Acad Sci U S A. 98:2509–2514.

Swofford DL. 2002. Phylogenetic analysis using parsimony (and othermethods) v4. Sunderland (MA): Sinauer Associates.

Terai Y, Seehausen O, Sasaki T, Takahashi K, Mizoiri S, Sugawara T, SatoT, Watanabe M, Konijnendijk N, Mrosso HDJ, et al. 2006. Divergentselection on opsins drives incipient speciation in Lake Victoria cich-lids. PLoS Biol. 4:e433.

Tyack PL, Johnson M, Soto NA, Sturlese A, Madsen PT. 2006. Extremediving of beaked whales. J Exp Biol. 209:4238–4253.

Uhen MD. 2010. The origin(s) of whales. Annu Rev Earth Planet Sci.38:189–219.

Van Nynatten A, Bloom D, Chang BSW, Lovejoy NR. 2015. Out of theblue: adaptive visual pigment evolution accompanies Amazon inva-sion. Biol Letters. 11(7):20150349.

Veilleux CC, Louis EE, Bolnick DA. 2013. Nocturnal light environ-ments influence color vision and signatures of selection onthe OPN1SW opsin gene in nocturnal lemurs. Mol Biol Evol.30:1420–1437.

Wald G. 1968. Molecular basis of visual excitation. Science 162:230–239.

Wang JY, Riehl KN, Dungan SZ. 2014. Family Delphinidae (ocean dol-phins). In: Wilson DE, Mittermeier RA, editors. Handbook of themammals of the world. Vol. 4: Sea mammals. Barcelona (Spain):Lynx Edicions. p. 410–527.

Warrant EJ, Johnsen S. 2013. Vision and the light environment. Curr Biol.23:R990–R994.

Warrant EJ, Locket NA. 2004. Vision in the deep sea. Biol Rev CambPhilos Soc. 79:671–712.

Watwood SL, Miller PJO, Johnson M, Madsen PT, Tyack PL. 2006. Deep-diving foraging behaviour of sperm whales (Physeter macrocepha-lus). J Anim Ecol. 75:814–825.

Weadick CJ, Chang BSW. 2012. An improved likelihood ratio test fordetecting site-specific functional divergence among clades of pro-tein-coding genes. Mol Biol Evol. 29:1297–1300.

Wiederstein M, Sippl MJ. 2007. ProSA-web: interactive web service forthe recognition of errors in three-dimensional structures of proteins.Nucleic Acids Res. 35:W407–W410.

Wozniak B, Dera J. 2007. Light absorption by suspended particulatematter (SPM) in sea water. In: Light absorption in sea water. NewYork: Springer. p. 167–294.

Yang Z. 2007. PAML 4: phylogenetic analysis by maximum likelihood.Mol Biol Evol. 24:1586–1591.

Yang Z, Wong WSW, Nielsen R. 2005. Bayes empirical Bayes inference ofamino acid sites under positive selection. Mol Biol Evol. 22:1107–1118.

Yim H-S, Cho YS, Guang X, Kang SG, Jeong J-Y, Cha S-S, Oh H-M, Lee J-H,Yang EC, Kwon KK, et al. 2014. Minke whale genome and aquaticadaptation in cetaceans. Nat Genet. 46:88–92.

Zhang J, Nielsen R, Yang Z. 2005. Evaluation of an improved branch-sitelikelihood method for detecting positive selection at the molecularlevel. Mol Biol Evol. 22:2472–2479.

Zhao H, Ru B, Teeling EC, Faulkes CG, Zhang S, Rossiter SJ. 2009.Rhodopsin molecular evolution in mammals inhabiting low lightenvironments. PLoS One 4:e8326.

Zhou X, Sun F, Xu S, Fan G, Zhu K, Liu X, Chen Y, Shi C, Yang Y, Huang Z,et al. 2013. Baiji genomes reveal low genetic variability and newinsights into secondary aquatic adaptations. Nat Commun. 4:2708.

14





ownloaded from


Documents

MBE Advance Access published November 24, 2015 ...chang.eeb.utoronto.ca/files/2016/02/Dungan-et-al-2016...Article Spectral Tuning of Killer Whale (Orcinus orca) Rhodopsin: Evidence